Multi-net Structures

The previously accepted way of designing a network architecture for a factory, was to have the Fieldbus directly connected to the sensors and actuators. The Fieldbus would then be connected to a cell-controller, and a number of cell-controllers would be then connected to a cell-network, and so on, up through the hierarchy, ending with a high speed backbone network. The data rate for the network on the next level up was assumed to be a magnitude higher than on the lower networks.

This was perhaps a reasonable philosophy in the past, where all data had to eventually end up in a powerful computer at the top level. The technique for today and the future,is to distribute intelligence between the cell-controllers, interfaces and sensors. At each level, the data becomes concentrated and regulating loops are typically closed within the same bus.

The need for a fast data rate at the higher levels is now decreasing, as more intelligence is distributed. This is the reason why P-NET may be used on several levels in a complete factory automation system.

Figure 3: A Multi-net structure with the P-NET Fieldbus

Dividing a system into cells, corresponding with each section of a plant, makes it possible to shut down a single section without affecting others. Program execution may be distributed in one or more independent processors per cell.

A software or hardware error in one cell, would not affect the others. An individual cell now only has a limited need to exchange data with other cells, e.g. to start and stop processes, to load recipes, to transfer production data etc.

In systems with real distributed intelligence, additional processing power can always be added in the form of additional master controllers. It is therefore possible for a system like this to be expanded.

Among the available Fieldbus systems, only P-NET allows direct addressing between several bus segments, also known as a multi-net structure. This feature is a specified part of the P-NET protocol, and it can be built into the standard operating system of multi-port masters. A multi-net structure is illustrated in fig. 3.

Communication is directed through the different bus segments via nodes with two or more P-NET interfaces. This means that any master on one bus segment can transparently access any node within any other bus segment, without the need for special programs in the multi-port masters. See fig. 4.

The segmentation also makes it possible to have independent local traffic on each bus segment, which increases the update rate and the data throughput throughout the total system.

Figure 4: Transparent access through multi-port masters to other bus segments.The benefits gained by dividing a system into smaller sections are highly significant, because it limits the consequence of an error, to a single segment, which gives higher system security. Furthermore, these multi-net features provide a natural redundancy, which makes the total plant installation very robust with respect to errors. See also fig. 3. An important advantage of the P-NET multi-net topology, is that there is no need for a hierarchical structuring of the bus segments. This is of great benefit when expanding existing P-NET installations, and when coupling to other networks.

An attempt to connect two segments within one node, using a bus system without this multi-net facility, requires a special program in that node. Such a program needs to collect all the data from all devices in one segment to make it available to the other segment, which is known as creating “process images”.

With the large amount of data that are available in today’s intelligent nodes, it is almost impossible to update and maintain a true “process image” for a complete bus segment. Such a procedure occupies a significant percentage of the bus capacity and requires a large amount of memory. Furthermore, it is expensive to create and test a dedicated program for each segment connection.
P-NET does not require such complex “process images” to be built.